**BETA BUILD** LTX-2 EASY PROMPT v2 + VISION Node
New workflow uses Qwen2.5-VL-3B to analyze images and generate scene-perfect animation prompts.
Deep Dive
Developer Wi released LTX-2's Easy Prompt v2 + Vision Node, an open-source ComfyUI workflow. It uses a local vision model (Qwen2.5-VL-3B) to analyze a starting image and generate a detailed scene context. This context is fed to an LLM to create precise animation prompts that maintain visual consistency. New features include automatic negative prompts, LoRA trigger word injection, and a dialogue toggle for better audio sync.
Why It Matters
Enables creators to generate consistent, high-quality AI video from a single reference image, significantly improving workflow efficiency.